In this paper (see path to it below), Michael Harris offers comments that compare trend following with other trading, all with respect to patterns. (see pages 5 and 14).
http://www.tradingpatterns.com/Literature/profitability.pdf
On page 5 he suggests higher trend testing requirements for Rwl (win loss ratio) and on page 14 he introduces an alpha to constrain failure by being conservative. The trend aspect and non trend aspect do not even have any overlap in the alpha.
This is also common to the thinking of Joe Ross and Covel in their respective documentation and speech giving.
To trace back to the common misconceptions represented, I have concluded that they made assumptiions based on a common distortion in their thinkng.
To profit from not making these mutual errors, it is important to include additional variables in any work done in terms of research, back testing, constructing models and doing model development. It is becoming clearer and clearer as to why managed money that involves trend following lacks a quality performance. The foundation is shakey, apparently.
The APS Automatic Pattern Search software was only used in a basic manner for writing the paper above. This limitation produced the paper's results up to the point of the examples.
The reference in the paper I point to could not have come from that context alone. They are very focussed extrapolations not seen in the examples at all.
There is no comparison between Lo and Harris. Lo choses patterns and evaluates trading them. Harris lets a coded system find patterns and then determines their viability using standard measures. Both perpetuate omitting using variables of the market.
Could this be an industry wide phenomena? If so, anyone using this type of work should be very cautious if not enhancing it to some minimum standard like using the variables of the market.
